文献速递神经技术与安全MR

基于特征金字塔网络的计算机辅助检测及监测脑转移瘤对比增强MRI治疗反应

Feature pyramid network-based computer-aided detection and monitoring treatment response of brain metastases on contrast-enhanced MRI.
PubMed 文献 · Clin Radiol · 2023-07-29

本研究回顾性纳入85例新诊断脑转移瘤患者,比较人工检测与基于特征金字塔网络的计算机辅助检测在初始和随访三维对比增强MRI上的表现。计算机辅助检测的敏感度接近100%,假阴性更少,且与人工检测在评估治疗反应方面一致性很好。

要点

  • 基于特征金字塔网络的计算机辅助检测对脑转移瘤的敏感度接近100%,明显高于人工检测。
  • 计算机辅助检测在随访MRI上敏感度仍高,假阴性少于人工检测。
  • 两位阅片者在评估脑转移瘤治疗反应方面一致性很好。

核心结论

  • 基于特征金字塔网络的计算机辅助检测对脑转移瘤检测敏感度接近100%、假阴性少于人工检测,在监测治疗反应方面表现良好,但在反映治疗后病灶变化方面存在少数不足。

研究设计与样本

  • 回顾性研究,纳入85例新诊断脑转移瘤并接受初始及随访三维对比增强MRI的癌症患者;人工检测由阅片者1完成,计算机辅助检测由阅片者2使用uAI Discover-BMs软件完成,两位阅片者分别评估每位患者的治疗反应。

诊断标准与关键数值

  • 初始三维对比增强MRI上人工检测敏感度为78.65%,计算机辅助检测敏感度为99.13%;随访三维对比增强MRI上人工检测敏感度为76.32%,计算机辅助检测敏感度为98.24%;采用配对卡方检验比较两种方法对脑转移瘤检出的差异,采用kappa检验评估两种方法监测治疗反应的一致性。

临床应用与阅片要点

  • 作者认为基于特征金字塔网络的计算机辅助检测在对比增强MRI上监测脑转移瘤治疗反应方面具有较高性能,尽管在反映治疗后脑转移瘤变化方面存在少数不足。

局限性

  • 计算机辅助检测在反映治疗后脑转移瘤变化方面存在少数不足。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

英文原文(摘要)· 点击展开

AIM: To investigate the value of feature pyramid network (FPN)-based computer-aided detection (CAD) of brain metastases (BMs) before and after non-surgical treatment, and to evaluate its performance in monitoring treatment response of BM on contrast-enhanced (CE) magnetic resonance imaging (MRI).

MATERIAL AND METHODS: Eighty-five cancer patients newly diagnosed with BM who had undergone initial and follow-up three-dimensional (3D) CE MRI at Liaocheng People's Hospital were included retrospectively in this study. Manual detection (MD) was performed by reviewer 1. Computer-aided detection (CAD) was performed by reviewer 2 using uAI Discover-BMs software. The treatment response was assessed by the two reviewers for each patient separately. A paired chi-square test was used to compare the differences in the detection of BM between MD and CAD. Agreement between MD and CAD in monitoring treatment response was assessed by kappa test.

RESULTS: The sensitivities of MD and CAD on initial 3D CE MRI were 78.65% and 99.13%, respectively. The sensitivities of MD and CAD on follow-up 3D CE MRI were 76.32% and 98.24%, respectively. There was a very good agreement between Reviewer 1 and Reviewer 2 in evaluating the treatment response of BM.

CONCLUSION: FPN-based CAD has a higher sensitivity of close to 100% and lower false negatives (FNs) for BM detection, compared to MD. Although CAD had a few shortcomings in reflecting changes of BMs after treatment, it had high performance in monitoring treatment response of BM on CE MRI.

来源:Clin Radiol,查看原文。版权归原作者所有。

引用

  1. Chen J, Meng L, Bu C, et al. Feature pyramid network-based computer-aided detection and monitoring treatment response of brain metastases on contrast-enhanced MRI. Clin Radiol. 2023. PMID: 37573242. doi:10.1016/j.crad.2023.07.009

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。